基于局部边缘预测的空谱联合高光谱图像无损压缩

发布时间：2018-08-19 17:47

【摘要】：利用高光谱图像丰富的边缘特性和很强的谱间结构相似性,提出一种基于局部边缘预测的空谱联合高光谱图像无损压缩方法。该方法利用谱间最小方差算法的编码框架,在原有谱内、谱间预测模式的基础上,增加了第三种"无预测"的预测模式,以更好地适应高光谱图像的相关特性。在谱内预测时,针对图像中普遍存在的局部斜边缘,将对角边缘检测引入到中值预测中,提出了改进的对角边缘预测算法。在谱间预测时,通过分析局部边缘存在时上下文的特点,提出简单有效的上下文选择策略,在此基础上,提出了基于局部边缘结构相似性的谱间预测算法,在上下文模板内自适应地选择最佳预测上下文进行谱间预测。实验结果表明,本文方法有效利用了图像的局部边缘特性,更好地去除了谱内和谱间的相关性,改善了预测性能,提高了无损压缩比。
[Abstract]:Based on the rich edge characteristics of hyperspectral images and the strong structural similarity between spectra, a novel lossless compression method for hyperspectral images based on local edge prediction is proposed. This method uses the coding framework of the minimum variance algorithm between spectra and adds a third prediction model without prediction to fit the correlation characteristics of hyperspectral images on the basis of the original spectral and interspectral prediction models. The diagonal edge detection is introduced into the median prediction for the local oblique edge which is widely existed in the image. An improved diagonal edge prediction algorithm is proposed. In the interspectral prediction, a simple and effective context selection strategy is proposed by analyzing the characteristics of the context when the local edge exists. On this basis, an inter-spectral prediction algorithm based on the similarity of the local edge structure is proposed. The optimal prediction context is chosen adaptively in the context template for interspectral prediction. The experimental results show that the proposed method can effectively utilize the local edge characteristics of the image, remove the correlation between spectra, improve the prediction performance and improve the lossless compression ratio.
【作者单位】：西安电子科技大学综合业务网国家重点实验室;西安电子科技大学软件学院;
【基金】：国家自然科学基金项目(61301291;61401324) 高等学校创新引智基地项目(B08038)
【分类号】：TP391.41

【相似文献】